* @brief aclnnMaxPool3dWithArgmax First segment interface. Calculate the workspace size based on the specific
* calculation process. Function description: Calculates the aclnnMaxPool3dWithArgmax from the input tensor, at the
* output contains 2 tensors: out and indices
* @domain aclnn_ops_infer
* @param [in] self: aclTensor on the NPU device. the data type can be float32/float16/bfloat16. The data format
* supports ND.
* @param [in] kernelSize: aclIntArray type, indicating the maxpooling window size.
* @param [in] stride: aclIntArray type, the step size of the window movement.
* @param [in] padding: aclIntArray type, number of padding layers for each edge. The value is negative infinity.
* @param [in] dilation: aclIntArray type, controls the stride of elements in the window.
* @param [in] ceilMode: bool type, when true, the output shape is calculated using round-up method. By default is
* rounding down.
* @param [in] out: aclTensor on the NPU device. the data type can be float32/float16/bfloat16. The data format supports
* ND.
* @param [in] indices: aclTensor on the NPU device. the data type can be int32. The data format supports ND.
* @param [out] workspaceSize: Returns the workspace size that the user needs to apply for on the npu device side.
* @param [out] executor: Return the op executor, including the operator calculation process.
* @return aclnnStatus: Return the status code.
*/
Thanks for sending an issue! Please fill in the following template to help quickly solve your problem.
Document Link（文档链接）
pooling\max_pool3d_with_argmax_v2\op_api\aclnn_max_pool3d_with_argmax.h
Issues Section（问题文档片段）
* @brief aclnnMaxPool3dWithArgmax First segment interface. Calculate the workspace size based on the specific * calculation process. Function description: Calculates the aclnnMaxPool3dWithArgmax from the input tensor, at the * output contains 2 tensors: out and indices * @domain aclnn_ops_infer * @param [in] self: aclTensor on the NPU device. the data type can be float32/float16/bfloat16. The data format * supports ND. * @param [in] kernelSize: aclIntArray type, indicating the maxpooling window size. * @param [in] stride: aclIntArray type, the step size of the window movement. * @param [in] padding: aclIntArray type, number of padding layers for each edge. The value is negative infinity. * @param [in] dilation: aclIntArray type, controls the stride of elements in the window. * @param [in] ceilMode: bool type, when true, the output shape is calculated using round-up method. By default is * rounding down. * @param [in] out: aclTensor on the NPU device. the data type can be float32/float16/bfloat16. The data format supports * ND. * @param [in] indices: aclTensor on the NPU device. the data type can be int32. The data format supports ND. * @param [out] workspaceSize: Returns the workspace size that the user needs to apply for on the npu device side. * @param [out] executor: Return the op executor, including the operator calculation process. * @return aclnnStatus: Return the status code. */Existing Issues（存在的问题）
aclnnMaxPool3dWithArgmax 已新支持int64索引，接口注释需更新